Lessons from standing up a forward deployed engineering practice inside a large enterprise, and how it fits next to the rest of engineering.
| 01Scrum Team of the Future | 02Forward Deployed Engineers | |
|---|---|---|
| Mission | Build the known roadmap, faster | Find and attack the unknown |
| Where | Inside an existing product team | Embedded with the business, where engineering is not |
| Shape | Smaller pods, same output | Small, senior, generalist |
| Horizon | Long lived ownership | Weeks. Scope, prove, hand off |
| Judged on | Throughput and quality | Business value delivered |
This is how scrum teams evolve with AI. The work does not change. The number of people it takes to deliver it does.
Five or more people, sized for the effort it used to take to write, test, and ship code.
Groups of three delivering the same result, with AI absorbing much of the build work.
FDEs are the first people in. They scout the lay of the land before anyone commits a roadmap, then pull out whichever tool the situation needs.
Sit with the business. Map the process, the systems, the data, and the people.
Separate the real problem from the requested one. Define what success looks like.
Working software in days, in front of real users. Not a slide.
Wire into the systems and data the business already runs on.
Tell the business when AI is the wrong answer, and what to do instead.
Package the work so a team can own it long after the FDE moves on.
Parts of the org with real, expensive problems and no engineering coverage. The work runs on spreadsheets, inboxes, and heroics.
There is no roadmap. Nobody has defined the product yet.
There is no backlog. The requirements do not exist until someone goes and finds them.
They create both. Discovery turns into a scoped problem, then a working prototype, then something a team can own.
Embed with the business. Map the process, systems, data, and decision makers.
Name the problem, the success metric, and the constraints. Baseline today's process.
Decide who should own it: FDE, a scrum team, or nobody.
Prototype on the paved road. Iterate with real users every week.
Report the delta against the baseline. Transition to an owning team.
Hours saved per week. Cycle time from request to result.
Cost avoided, revenue protected, spend consolidated.
Throughput per person. Manual steps removed.
Active users. Old workflows actually retired.
It runs on a paved road: an internal platform for rapid prototyping that takes an FDE from idea to a running demo in days.
Getting through enterprise red tape once, at the platform level, so every engagement inherits the approval instead of re-fighting it.
An internal platform that lets the practice compound. What one FDE learns is available to the next.
Scout a desert and find the real problem.
Prototype and measure value against the baseline.
A pod of three takes ownership and hardens it.
Running it in production reveals the next opportunity.